--- title: "Together AI Pricing 2026: Tokens, Endpoints, GPU Hours" url: "https://dupple.com/learn/together-ai-pricing" description: "Together AI pricing in 2026: serverless from $0.14 per 1M tokens, dedicated H100 endpoints at $5.49 an hour, GPU clusters from $3.99, monthly costs worked out." published: "2026-11-14" updated: "2026-11-14" llms_txt: "https://dupple.com/llms.txt" --- # Together AI Pricing 2026: Serverless Tokens, Dedicated Endpoints and GPU Hours Trusted by 800,000+ Techpresso subscribers · Editorial policy · How we make money Written by Louis Corneloup Founder at Dupple, covering AI tools and strategies for 800K+ readers. How our guides are researched . November 14, 2026 · Updated November 2026 8 min read **Key facts6 products compared: Serverless inference, Batch inference, Dedicated model inference, GPU…** - **Updated:** November 14, 2026 - **Prices as of:** November 2026 - 6 products compared: Serverless inference, Batch inference, Dedicated model inference, GPU clusters, Fine-tuning, Provisioned throughput - Serverless inference (best for: Apps with low or bursty traffic): $0.14 per 1M tokens (Llama 3 8B Instruct Lite) - Batch inference (best for: Offline jobs that can wait up to 24 hours): Varies by model - Dedicated model inference (best for: Steady, high-volume traffic and models not on serverless): $5.49 per H100 hour - GPU clusters (best for: Training and teams running their own serving stack): $1.99 per H100 hour preemptible, $3.99 on demand Short answer: Together AI is prepaid and usage-based, with no subscription. **Serverless models bill per million tokens**, from $0.15 input and $0.60 output on gpt-oss-120B up to $3.00 and $15.00 on Kimi K3. A dedicated H100 endpoint costs $5.49 per GPU-hour, and raw H100s in a GPU cluster cost $3.99 an hour on demand. You need a $15 first credit purchase to start. Checked on Together's pricing page, 30 October 2026. The pricing question most teams actually face is serverless vs dedicated: below roughly a billion tokens a month, per-token billing almost always wins, and above that the answer depends on which model you run. The scenario tables below show where the line falls. ## Together AI pricing at a glance *Together AI pricing at a glance: Together AI Pricing 2026: Serverless Tokens, Dedicated Endpoints and GPU Hours. Last updated November 2026.* | Product | How it bills | Starting price | Best for | | --- | --- | --- | --- | | Serverless inference | Per 1M input and output tokens | $0.14 per 1M tokens (Llama 3 8B Instruct Lite) | Apps with low or bursty traffic | | Batch inference | Per token, up to 50% off serverless | Varies by model | Offline jobs that can wait up to 24 hours | | Dedicated model inference | Per GPU-hour, billed by the minute | $5.49 per H100 hour | Steady, high-volume traffic and models not on serverless | | GPU clusters | Per GPU-hour | $1.99 per H100 hour preemptible, $3.99 on demand | Training and teams running their own serving stack | | Fine-tuning | Per 1M training tokens, with a per-job minimum | $0.34 per 1M tokens on the smallest models | Adapting an open model to your data | | Provisioned throughput | Per throughput unit | Estimate on the pricing page | Token capacity with an SLA | Prices from [Together AI pricing](https://www.together.ai/pricing), [dedicated inference pricing docs](https://docs.together.ai/docs/dedicated-endpoints/pricing) and [batch docs](https://docs.together.ai/docs/inference/batch/overview), read on 30 October 2026. Together has no free trial. The [credits docs](https://docs.together.ai/docs/billing-credits) say the first purchase must be at least $15 and later top-ups at least $5, credits do not expire, and API access stops when the balance hits zero. That prepaid model is a real safety feature for a side project: you cannot wake up to a surprise bill. ## How much do serverless models cost on Together AI? Output prices vary about 100-fold between models, so the model you pick is the budget. The page even lists one model, Ternary Bonsai 27B, at $0.00. A selection from the pricing page: *How much do serverless models cost on Together AI?: Together AI Pricing 2026: Serverless Tokens, Dedicated Endpoints and GPU Hours. Last updated November 2026.* | Model | Input per 1M tokens | Cached input per 1M | Output per 1M tokens | | --- | --- | --- | --- | | Llama 3 8B Instruct Lite | $0.14 | Not listed | $0.14 | | gpt-oss-120B | $0.15 | Not listed | $0.60 | | GLM-5.3-Flash | $0.15 | $0.03 | $0.50 | | DeepSeek V4.1 Flash | $0.30 | $0.006 | $1.20 | | Qwen3.5-397B-A17B | $0.60 | $0.35 | $3.60 | | GLM-5.3 | $1.40 | $0.26 | $4.40 | | Qwen3.8-2.4T-A95B | $2.00 | $0.25 | $6.00 | | Kimi K3 | $3.00 | $0.30 | $15.00 | If tokens are new to you, [what a token is in AI](https://dupple.com/learn/what-is-a-token-in-ai) explains the unit. Output tokens cost three to six times as much as input on most large models, so a chatbot that writes long answers costs far more than a classifier that returns one word. Cached input is the other lever: on DeepSeek V4.1 Flash a cached token costs $0.006 per million against $0.30 uncached, so long system prompts that repeat on every call get much cheaper. One change that catches people out: **Llama 3.3 70B is no longer on Together's serverless API.** Its model page says so, and the pricing page lists no 70B Llama for serverless. You can still run it, but only on a dedicated endpoint billed by the GPU-hour. ## What does a dedicated endpoint cost per month? Dedicated model inference bills for hardware, not tokens. An H100 80GB costs $5.49 an hour and a B200 180GB costs $8.99 an hour, both on demand; H200, B300 and GB300 are quote-only. Larger models need more GPUs, and Together's docs say a four-GPU configuration costs four times the single-GPU rate. *What does a dedicated endpoint cost per month?: Together AI Pricing 2026: Serverless Tokens, Dedicated Endpoints and GPU Hours. Last updated November 2026.* | Usage pattern | H100 replica | B200 replica | | --- | --- | --- | | Always on, 30-day month (720 hours) | $3,952.80 | $6,472.80 | | 8 hours a day, 22 working days | $966.24 | $1,582.24 | All figures are our arithmetic: hourly rate × hours, per replica, before any reserved discount. Billing runs by the minute, and only while a replica is ready to serve. Provisioning and cold starts are not billed, and a deployment scaled to zero costs nothing. That makes the second row realistic for internal tools: set an inactivity timeout and the endpoint stops itself overnight. Reserved capacity, at a lower effective rate for a committed term, is quote-only. **Dupple data:** the same H100 costs 38% more per hour as a managed dedicated-inference replica ($5.49) than as an on-demand cluster GPU ($3.99), our arithmetic from Together's pricing page on 30 October 2026. The premium buys autoscaling, traffic routing and a ready inference API; a cluster gives you bare GPUs and you run the serving stack yourself. ## Serverless or dedicated: which is cheaper at 50M and 500M tokens? Here is the monthly serverless bill at two volumes, assuming three input tokens for every output token, a common shape for chat and retrieval apps. Compare each figure with $3,952.80 for one always-on H100 replica. *Serverless or dedicated: which is cheaper at 50M and 500M tokens?: Together AI Pricing 2026: Serverless Tokens, Dedicated Endpoints and GPU Hours. Last updated November 2026.* | Model | 50M tokens a month | 500M tokens a month | Volume where one H100 replica costs the same | | --- | --- | --- | --- | | gpt-oss-120B | $13.13 | $131.25 | About 15 billion tokens | | GLM-5.3 | $107.50 | $1,075.00 | About 1.8 billion tokens | | Kimi K3 | $300.00 | $3,000.00 | About 660 million tokens | Our arithmetic: 75% of tokens at the input price, 25% at the output price, no caching, no batch discount. The answer is lopsided. At 50M tokens a month, serverless is cheaper than a dedicated H100 for every model on the list, by a wide margin. At 500M, only Kimi K3 comes close to the cost of a replica, and only because its output price is high. Dedicated still makes sense in three cases. The model you need is not on serverless, as with Llama 3.3 70B. You need guaranteed latency and no shared rate limits. Or your traffic is so steady that a replica stays busy all day, which Together's own docs name as the point where dedicated usually wins. Before you commit, check whether one H100 can actually carry the model: a model that needs two or four GPUs doubles or quadruples the replica cost. ## How much does batch inference save? Batch jobs run from one uploaded JSONL file and return results within a 24-hour window, at **up to 50% off serverless rates**, with a separate rate-limit pool. On the GLM-5.3 example above, that would take a 500M-token month from $1,075 to as little as $537.50 if the model qualifies for the full discount (our arithmetic). The trade-off is latency. Batch suits classification, evaluations, synthetic data and overnight summaries, not anything a user is waiting on. Together keeps batch input and output files for 7 days, so download results before they expire. ## What do GPU clusters cost? GPU clusters rent raw GPUs by the hour, with longer commitments bringing the rate down: *What do GPU clusters cost?: Together AI Pricing 2026: Serverless Tokens, Dedicated Endpoints and GPU Hours. Last updated November 2026.* | GPU | Preemptible | On demand | Reserved 3 to 30 days | 31 to 90 days | 91 to 180 days | | --- | --- | --- | --- | --- | --- | | NVIDIA HGX H100 | $1.99 | $3.99 | $3.69 | $3.45 | $3.19 | | NVIDIA HGX H200 | $2.99 | $5.99 | $4.99 | $4.15 | $3.99 | | NVIDIA HGX B200 | $4.09 | $8.19 | $7.99 | $7.79 | $6.79 | | NVIDIA HGX B300 | $4.99 | $9.99 | Contact us | Contact us | Contact us | Prices per GPU per hour from [Together AI pricing](https://www.together.ai/pricing), same date. Terms of 181 days or more, and GB200 and GB300 NVL72 systems, are quote-only. Preemptible GPUs cost half the on-demand rate but can be taken back, so they fit fault-tolerant training with checkpoints, not serving. Shared filesystem storage alongside a cluster costs $0.16 per GiB a month. For how these rates compare with other GPU clouds, Devshot keeps a [cheapest GPU cloud comparison](https://devshot.co/blog/cheapest-gpu-cloud). ## What does fine-tuning cost? Fine-tuning bills per million tokens processed: training set size × epochs, plus any evaluation tokens. LoRA supervised fine-tuning starts at $0.34 per 1M tokens on small models such as Llama 3.1 8B and Qwen3.5 9B, and Llama 3.3 70B costs $2.03 per 1M. Each job carries a minimum charge, $4 on most small models and up to $60 on the largest. Worked example: a 20M-token dataset trained for 3 epochs on Llama 3.3 70B is 60M tokens, or $121.80 with LoRA SFT (our arithmetic). Serving the result then needs a dedicated endpoint, since the base model is not on serverless. Together also sells post-training compute by the GPU-hour. B200 time is on a promotion at $9.00 an hour instead of $11.00 until 14 December 2026, so budget for the higher figure on any run after that date. ## The costs that surprise new users - **No free tier.** There is no trial; the $15 first purchase is the entry fee. - **Idle replicas bill.** A dedicated endpoint with a minimum replica count above zero bills around the clock, traffic or not. - **Multi-GPU models multiply.** A model that needs four H100s costs four times the single-GPU rate per replica. - **Models leave serverless.** Llama 3.3 70B moved off the serverless API, so a model you build on today may need a dedicated endpoint later. - **Displayed image and video prices are minimums.** The pricing page says they refer to the lowest resolution and duration settings. If you are comparing providers rather than models, a router can send each request to the cheapest host for that model: see our list of [LLM gateways](https://dupple.com/learn/best-llm-gateways). Running small models on your own machine costs nothing per token, and [the best free local LLM tools](https://dupple.com/learn/best-free-local-llm-tools) cover that route. For the apps that sit on top, [the best RAG tools](https://dupple.com/learn/best-rag-tools) and [the best AI agent frameworks](https://dupple.com/learn/best-ai-agent-frameworks) are the next stop, and [OpenRouter pricing](https://dupple.com/learn/openrouter-pricing) breaks down the main multi-provider alternative. ## How we checked these prices We read [Together AI's pricing page](https://www.together.ai/pricing), the [Llama 3.3 70B model page](https://www.together.ai/models/llama-3-3-70b), and the docs on [dedicated inference pricing](https://docs.together.ai/docs/dedicated-endpoints/pricing), [batch inference](https://docs.together.ai/docs/inference/batch/overview) and [credits](https://docs.together.ai/docs/billing-credits) on 30 October 2026. Scenario volumes, the 3:1 input-to-output split and every monthly total are our own arithmetic from list prices. Together's model list changes often, so check the model you plan to use on the pricing page before you budget. A summary of the same plans sits on [Together AI's Toolradar pricing profile](https://toolradar.com/tools/together-ai/pricing). ## FAQ **How much does Together AI cost in 2026?** Together AI bills by usage with no subscription. Serverless models range from $0.14 per 1M tokens on Llama 3 8B Instruct Lite to $3.00 input and $15.00 output on Kimi K3, dedicated H100 endpoints cost $5.49 per GPU-hour, and on-demand H100 cluster GPUs cost $3.99 an hour. The first credit purchase must be at least $15. **Does Together AI have a free tier?** No. Together's docs say it offers no free trial and requires a minimum $15 credit purchase before you can use the platform. Credits do not expire. **Is Together AI cheaper than running a dedicated GPU?** For most workloads under about a billion tokens a month, yes. At 500M tokens a month, GLM-5.3 on serverless costs about $1,075 against $3,952.80 for one always-on H100 replica (our arithmetic). Dedicated wins when a replica stays busy all day or the model is not on serverless. **Can I run Llama 3.3 70B on Together AI?** Yes, but only on a dedicated endpoint. Its model page says it is not available on Together's serverless API, so you pay per GPU-hour instead of per token. **How does Together AI bill dedicated endpoints?** By the minute, per running replica, at the hardware's hourly rate, which is higher for a B200 than an H100 and multiplies with GPU count. Provisioning, cold starts and replicas scaled to zero are not billed. **What is Together AI's batch discount?** Batch jobs cost up to 50% less than serverless rates, use a separate rate-limit pool, and complete within a 24-hour window. They suit offline work such as classification and evaluations. ## Bottom line Start on serverless, pick the cheapest model that does the job, and use batch for anything that can wait; [how to reduce LLM API costs](https://dupple.com/learn/how-to-reduce-llm-api-costs) lists the other levers. Move to a dedicated endpoint only when the model you need is not on serverless or your traffic keeps a replica busy most of the day, and run the per-month arithmetic on the exact GPU count first. **Cite this:** Dupple, "Together AI Pricing 2026: Serverless Tokens, Dedicated Endpoints and GPU Hours", November 2026. **Pricing out the AI stack for your team?** [Start the Dupple X trial](https://checkout.dupple.com/checkout/dupple-x-yearly-trial/). Some offers on this page may be paid placements or contain affiliate links. Related Articles Blog Post ### Best Free Serverless Platforms in 2026 The best free serverless platforms in 2026, with verified prices and aggregated ratings. Cloudflare Workers starts at $5, Vercel at $20 and Render at $25, and the free tier is not where the cost lives. Article ### How Much Does It Cost to Fine-Tune an LLM in 2026? LoRA vs Full, GPU vs API What fine-tuning an LLM costs in 2026: hosted LoRA runs from $10 on 8B models, 70B runs from $61, GPU rental from $1.99 an hour, worked out from vendor prices. Article ### Claude Code Pricing 2026: Every Plan, API Rates and Real Monthly Cost Claude Code pricing in 2026: included in Pro ($17 or $20/mo), Max ($100 or $200/mo) and Team seats, or pay per token on the API. Real monthly cost worked out. Blog Post ### Best AI Model Training Tools in 2026 I compared the best AI model training tools of 2026, from Unsloth and Hugging Face AutoTrain to Together AI, Modal, and Weights & Biases. Real pricing, honest picks. Article ### Claude Code vs Cursor in 2026: Which AI Coding Tool Should You Use? Claude Code vs Cursor in 2026: pricing, Claude Opus 5.5 and Sonnet 5, IDE integration, multi-file refactoring, and which to pick. Plus how devs use both together. Article ### Is DeepSeek Free in 2026? Chat, API and Self-Hosting Costs DeepSeek chat lists no paid plan, the API bills per token from $0.15 per million input tokens off-peak, and the MIT-licensed models are free to self-host.